Challenge
The bank wished to obtain a system that will automatically calculate the risks regarding fundings and credit. We proposed further improvement for automated data analysis, implementing an AI document analysis system primarily focused on data extraction.
About the Project
The client is one of the enterprise financial services providers. It deals with a vast range of services, including corporate, retail, investment and private banking along with asset management.
Region
USA
Time
10 months
Team
5 specialists
The technologies we used
Considering the client request we decided to promote Proof-of-Concept for Intellectual Document Processing solution. The solution itself is based on the principle of intelligent document analysis which is further connected with automated data processing.
Before we get to the detailed presentation of the solution we presented, let’s say a few words about these two principles.
Sample doc source
For business purposes IDP allows, for instance, to extract relevant data, such as dates, names, numbers and amounts from the files, scans or photos sent by the clients. The processing itself provides all the necessary information for business, making it convenient for further analysis.
The main tool for that is optical character recognition (OCR) mechanism and this mechanism requires additional overview.
When you receive a scan, photo or PDF, the OCR program finds the coordinates of the text parts which should be processed. It also defines whether the text is presented in the form of table, formula, etc. When all the characters or symbols are analyzed, the IDP identifies their semantics. In other words, the AI algorithm understands and notifies that some of the symbols or even the whole words can be someone’s address, phone number or name, or the subsequence of numbers can be a sum of money or bank account number.
In the end, answering the question “What is document processing?”, we say that it provides the business using it with the cluster of structured data which in most cases is transferred into its database.
The main benefit of using such a form of document analysis is that the program does the entire work for you. Instead of dealing with a team of employees who will read through everything that the company receives, it provides automated data extraction from the document, highlighting the key points and making them ready for further analysis.
What is automated data processing?
IDP, however, is only the first step to decision-making optimization. It goes on with the second principle.
Automated data processing (ADP) is based on usage of technologies, including AI, which analyze structured data to get necessary information.
For instance, through ADP business can get some insights or detailed reports and spreadsheets that will help to make the decision-making faster and easier. As an example, we’ve had a solution like this, referring to AI predictive analytics, that allowed our clients to decide which banner will be the most attractive for their website visitors.
In our case, the bank had the opportunity to see whether the client is reliable for credit or not, as they could get the information about his credit story, incomes and obligatory payments, such as taxes and utility bills.
Our solution
Getting acquainted with the client's needs, we offered a solution that will help to get rid of unnecessary work connected with time-consuming document analysis.
We had already created a similar solution in the past providing a tool for AI resume screening.
- 1
The first problem that occurred is the task of data gathering. In typical projects of AI document extraction, the work starts from gathering the necessary document samples, which should be analyzed in order to train the AI system. This process is rather painstaking and requires working with hundreds of documents at least. Moreover, this amount of work can’t present any immediate results, as this is only the first, initial stage of development.
We’re able to spare the client from it. Our principle was to take several documents of different types (the most popular in the client’s area) and find the key points and value types that will define the analyzing process of all the other information that the bank clients would bring.
Through that, we optimized the initial and the most unpleasant part of the work as it now could take only a few seconds to perform. The same solution was invented before for contract analysis – the task that companies face every single day.
- 2
Then, when the first step was passed, we used the document processing AI principle based on OCR technology to efficiently process tons of documents the bank receives every day. Two main tasks that were covered were document classification and automated data extraction.
In the end, the bank obtained the software which it lacked so much. It was a link between the unstructured client documents and their efficient and productive automated analysis.
Conclusion
A document processing platform is certainly a must-have tool for any business and for the banking sphere in particular. It’s just one of many FinTech solutions that we’ve implemented allowing businesses to automize the processes of decision making through efficient automated document analysis.
Its implementation along with automated data processing programs is necessary to calculate the funding or lending risks with the minimum of money, time and working costs.
Our Expert
"The main goal of our solution is to spare specialists from financial and other industries from the need to spend hours manually analyzing and processing documents. With our AI tool, you can automatically extract all the necessary useful information from documents, quickly assess credit risks, automatically extract and classify data from documents, while reducing time and labor costs."
Nikolai Karelin - Head of AI
Frequently Asked Questions
Document AI is a more advanced approach that combines OCR with NLP and ML technologies. At the first stages document AI also uses OCR methods to identify and extract text from the document. Then, however, ML and NLP tools start their work, providing greater functionality. For example, document AI can be focused on the context and text semantics through linking company or person’s name with their address or phone number. It also can differentiate document entities such as people, organizations, locations and organize them. All the data extracted and processed within the document AI can be automatically transformed into databases or structured spreadsheets.
IDP is a universal tool that works in every business which uses documents in any way. From real estate to healthcare, from transport to commercial market – IDP will be a perfect solution for document processing optimization.
Yes, AI can be used to scan and analyze documents, but the specific capabilities depend on the type of AI and the tools used. AI can perform document classification, data summarization or even translate the text and its parts.
Project info
